Autonomous vehicles play a key role in the smart cities vision: they bring benefits and innovation, but also safety threats, especially if they suffer from vulnerabilities that can be easily exploited. In this paper, we propose a method that exploits Deep Reinforcement Learning to train autonomous vehicles with the purpose of preventing road accidents. The experimental results demonstrated that a single self-driving vehicle can help to optimise traffic flows and mitigate the number of collisions that would occur if there were no self-driving vehicles in the road network. Our results proved that the training progress is able to reduce the collision frequency from 1 collision every 32.40 hours to 1 collision every 53.55 hours, demonstrating the effectiveness of deep reinforcement learning in road accident prevention in smart cities.
A Method for Road Accident Prevention in Smart Cities based on Deep Reinforcement Learning
Mercaldo F.;Santone A.
2022-01-01
Abstract
Autonomous vehicles play a key role in the smart cities vision: they bring benefits and innovation, but also safety threats, especially if they suffer from vulnerabilities that can be easily exploited. In this paper, we propose a method that exploits Deep Reinforcement Learning to train autonomous vehicles with the purpose of preventing road accidents. The experimental results demonstrated that a single self-driving vehicle can help to optimise traffic flows and mitigate the number of collisions that would occur if there were no self-driving vehicles in the road network. Our results proved that the training progress is able to reduce the collision frequency from 1 collision every 32.40 hours to 1 collision every 53.55 hours, demonstrating the effectiveness of deep reinforcement learning in road accident prevention in smart cities.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
